feat: 维修时间f1>80

core/tools/segment_llm.py

@@ -12,7 +12,9 @@ from core.config import LLM, MAX_WORKERS
 class LLMTool(ToolBase):
 class LLMTool(ToolBase):
     """LLM-backed processor: builds prompts, calls LLM, parses JSON."""
     """LLM-backed processor: builds prompts, calls LLM, parses JSON."""
 
 
-    def __init__(self, system_prompt: str, llm_key: str = "fastgpt_segment_review") -> None:
+    def __init__(
+        self, system_prompt: str, llm_key: str = "fastgpt_segment_review"
+    ) -> None:
         super().__init__()
         super().__init__()
         self.system_prompt = system_prompt
         self.system_prompt = system_prompt
         self.llm = OpenAITool(LLM[llm_key], max_workers=MAX_WORKERS)
         self.llm = OpenAITool(LLM[llm_key], max_workers=MAX_WORKERS)
@@ -33,7 +35,9 @@ class LLMTool(ToolBase):
         try:
         try:
             return asyncio.run(coro)
             return asyncio.run(coro)
         except RuntimeError as e:
         except RuntimeError as e:
-            print(f'RuntimeError in run_with_loop: {e}, trying to get event loop and run until complete.')
+            print(
+                f"RuntimeError in run_with_loop: {e}, trying to get event loop and run until complete."
+            )
             loop = asyncio.get_event_loop()
             loop = asyncio.get_event_loop()
             return loop.run_until_complete(coro)
             return loop.run_until_complete(coro)
 
 

data/batch/batch.py

@@ -13,30 +13,35 @@ from loguru import logger
 from utils.common_util import random_str
 from utils.common_util import random_str
 from utils.http_util import upload_file, fastgpt_openai_chat, download_file
 from utils.http_util import upload_file, fastgpt_openai_chat, download_file
 
 
-SUFFIX = "_麓发迁移"
-batch_input_dir_path = "jp-input"
-batch_output_dir_path = f"/home/ccran/lufa-contract/data/benchmark/results/jp-output-lufa-{time.strftime('%Y%m%d-%H%M%S', time.localtime())}"
+use_lufa = False
+
+if not use_lufa:
+    SUFFIX = "_麓发迁移"
+    batch_input_dir_path = "jp-input"
+    batch_output_dir_path = f"/home/ccran/lufa-contract/data/benchmark/results/jp-output-lufa-{time.strftime('%Y%m%d-%H%M%S', time.localtime())}"
+    # 金盘fastgpt接口
+    url = "http://192.168.252.71:18088/api/v1/chat/completions"
+    # 金盘迁移麓发合同审查测试token
+    token = "fastgpt-vykT6qs07g7hR4tL2MNJE6DdNCIxaQjEu3Cxw9nuTBFg8MAG3CkByvnXKxSNEyMK7"
+    # 人机交互测试（测试环境）
+    # token = 'fastgpt-p189K5zoTX5wjp0dBybFCwsbWm3juIwlJxt2wTGyiaOWOANI5Y10pKEZzyt'
+    # 人机交互测试（生产环境）
+    # token = 'fastgpt-ry4jIjgNwmNgufMr5jR0ncvJVmSS4GZl4bx2ItsNPoncdQzW9Na3IP1Xrankr'
+    # 提取后审查测试
+    # token = 'fastgpt-n74gGX5ZqLT6o1ysMBSGUTjIciswYOWDRfQ75krMkE5gDVDkpzsbz8u'
+else:
+    SUFFIX = "_麓发"
+    batch_input_dir_path = "lufa-input"
+    batch_output_dir_path = "lufa-output-standard"
+    # 麓发fastgpt接口
+    url = "http://192.168.252.71:18089/api/v1/chat/completions"
+    # 麓发合同审查生产token
+    # token = "fastgpt-ek3Z6PxI6sXgYc0jxzZ5bVGqrxwM6aVyfSmA6JVErJYBMr2KmYxrHwEUOIMSYz"
+    # 麓发合同审查生产token-标准化
+    token = "fastgpt-mg5tQUgreJeF7peoOr5zqP0NR4EIrfS2bEVXge6FUL94Suu1TvEMR1sGNRSiV"
+
 
 
-# SUFFIX = "_麓发"
-# batch_input_dir_path = "lufa-input"
-# batch_output_dir_path = "lufa-output-standard"
 batch_size = 5
 batch_size = 5
-# 麓发fastgpt接口
-# url = "http://192.168.252.71:18089/api/v1/chat/completions"
-# 金盘fastgpt接口
-url = "http://192.168.252.71:18088/api/v1/chat/completions"
-# 麓发合同审查生产token
-# token = "fastgpt-ek3Z6PxI6sXgYc0jxzZ5bVGqrxwM6aVyfSmA6JVErJYBMr2KmYxrHwEUOIMSYz"
-# 麓发合同审查生产token-标准化
-# token = "fastgpt-mg5tQUgreJeF7peoOr5zqP0NR4EIrfS2bEVXge6FUL94Suu1TvEMR1sGNRSiV"
-# 金盘迁移麓发合同审查测试token
-token = "fastgpt-vykT6qs07g7hR4tL2MNJE6DdNCIxaQjEu3Cxw9nuTBFg8MAG3CkByvnXKxSNEyMK7"
-# 人机交互测试（测试环境）
-# token = 'fastgpt-p189K5zoTX5wjp0dBybFCwsbWm3juIwlJxt2wTGyiaOWOANI5Y10pKEZzyt'
-# 人机交互测试（生产环境）
-# token = 'fastgpt-ry4jIjgNwmNgufMr5jR0ncvJVmSS4GZl4bx2ItsNPoncdQzW9Na3IP1Xrankr'
-# 提取后审查测试
-# token = 'fastgpt-n74gGX5ZqLT6o1ysMBSGUTjIciswYOWDRfQ75krMkE5gDVDkpzsbz8u'
 
 
 
 
 def extract_url(text):
 def extract_url(text):
@@ -94,10 +99,16 @@ def process_single_file(
         excel_url, doc_url = extract_url(result)
         excel_url, doc_url = extract_url(result)
         if excel_url and doc_url:
         if excel_url and doc_url:
             download_file(
             download_file(
-                excel_url.replace("218.77.58.8", "192.168.252.71"), des_excel_file
+                excel_url.replace("218.77.58.8", "192.168.252.71").replace(
+                    "znkf.lgfzgroup.com", "192.168.252.71"
+                ),
+                des_excel_file,
             )
             )
             download_file(
             download_file(
-                doc_url.replace("218.77.58.8", "192.168.252.71"), des_doc_file
+                doc_url.replace("218.77.58.8", "192.168.252.71").replace(
+                    "znkf.lgfzgroup.com", "192.168.252.71"
+                ),
+                des_doc_file,
             )
             )
             logger.info(
             logger.info(
                 f"第{counter}个文件下载:{excel_url}到{des_excel_file} {des_doc_file}"
                 f"第{counter}个文件下载:{excel_url}到{des_excel_file} {des_doc_file}"

data/benchmark/compare_annotation.py

@@ -24,8 +24,19 @@ def _load_rows(path: Path) -> list[tuple[str, str]]:
     for col in first_two.columns:
     for col in first_two.columns:
         first_two[col] = first_two[col].map(_normalize_cell)
         first_two[col] = first_two[col].map(_normalize_cell)
     first_two = first_two.replace("", pd.NA).dropna(how="all")
     first_two = first_two.replace("", pd.NA).dropna(how="all")
+    rows: list[tuple[str, str]] = []
+    seen: set[tuple[str, str]] = set()
+    for item, text in first_two.itertuples(index=False, name=None):
+        row = (
+            "" if pd.isna(item) else str(item).strip(),
+            "" if pd.isna(text) else str(text).strip(),
+        )
+        if row in seen:
+            continue
+        seen.add(row)
+        rows.append(row)
 
 
-    return list(map(tuple, first_two.itertuples(index=False, name=None)))
+    return rows
 
 
 
 
 def _compare_impl(val_dir: Path, answer_dir: Path) -> None:
 def _compare_impl(val_dir: Path, answer_dir: Path) -> None:

data/benchmark/eval.py

@@ -121,7 +121,7 @@ def _parse_args() -> argparse.Namespace:
     parser.add_argument(
     parser.add_argument(
         "--datasets-dir",
         "--datasets-dir",
         type=Path,
         type=Path,
-        default=base / "results" / "jp-output-lufa-20260408-182708",
+        default=base / "results" / "jp-output-lufa-20260416-000112",
         help="Directory containing Word files with annotations.",
         help="Directory containing Word files with annotations.",
     )
     )
     parser.add_argument(
     parser.add_argument(

main.py

@@ -394,9 +394,7 @@ def merge_segment_findings(payload: MergerRequest) -> MergerResponse:
     unqualified_findings = [
     unqualified_findings = [
         f for f in segment_findings if (f.result or "").strip() == "不合格"
         f for f in segment_findings if (f.result or "").strip() == "不合格"
     ]
     ]
-    merged_result = merger_tool.run(
-        [f.__dict__ for f in unqualified_findings], merge_mode="rule"
-    )
+    merged_result = merger_tool.run([f.__dict__ for f in unqualified_findings])
     merged_findings = merged_result.get("findings", []) or []
     merged_findings = merged_result.get("findings", []) or []
 
 
     for f in merged_findings:
     for f in merged_findings: